What automatic underwriting routing with an AI layer means
Automatic underwriting routing with AI is the stage where an external AI layer reads each incoming submission, scores it against the insurer's risk appetite and exposure bands, and sends it down the right path automatically: an instant quote, an automatic decline, or escalation to the right human subscritor (underwriter). It sits on top of the insurer's existing core and policy systems and connects through integration, so the core stays exactly where it is. This guide is written for the underwriting lead or innovation head who wants triage to be consistent and fast without running a core migration.
In Seguros e Danos (P&C), every case that arrives from the corretor (broker) channel has to be triaged before any pricing happens. Is the risk inside appetite, which ramo (line of business) and which authority level does it belong to, who should own it, and how urgent is it. When that triage is manual, the same failure modes recur. Cases land in the wrong queue and have to be re-routed, adding days. Work is picked up roughly first-in-first-out rather than by exposure or win probability, so high-value business waits behind noise. The corretor cannot tell whether a quote will come back in an hour or in three days. And two underwriters can read the same case and reach different appetite calls, which is both a conversion problem and a governance problem.
This matters in Brazil because volume keeps climbing. The Seguros e Danos market grows double digits per year, but company structure does not keep pace with that acceleration, so triage becomes a real bottleneck in the highest-volume lines. The pressure is well documented. Deloitte finds that underwriters spend 40% of their time on administrative tasks rather than risk judgment, and according to BCG, 70% of insurers do not execute innovation because of IT limitations. Automatic routing is the decision-and-prioritization stage that turns that pressure into consistent, auditable throughput, and it is exactly what WIR delivers as the AI layer of insurance.
How routing by appetite and exposure works
Automatic routing is the final stage of a longer automated subscrição (underwriting) journey, and it only works because the stages before it have already structured the case. The pipeline runs in six stages. First, multichannel intake captures submissions from API, portal, and upload in whatever format the insurer and its corretores already use, registering each one with a timestamp that starts the SLA clock. Second, intelligent document reading extracts structured fields from proposals, schedules of values, and prior loss runs, which removes the re-keying that consumes underwriter time. Third, broker enrichment adds context and a score by cross-referencing external and internal sources such as CNPJ validation, prior policy history, and exposure data. Fourth, a risk and fraud engine, a multi-factor Machine Learning model calibrated to appetite and the underwriting manual, produces a risk score and flags anomalies. Fifth, dynamic pricing calculates a risk-adjusted premium (prêmio) within the parameters the insurer defines.
The sixth stage is decision and prioritization, and it is where the AI layer earns its place. Each submission is scored on two axes at once. The first axis asks whether the risk is inside the defined risk appetite (apetite de risco) for that ramo. The second axis asks where it falls in the exposure and authority band, meaning sum insured, line size, and accumulation. The combination of those two axes determines the path. A risk clearly inside appetite and inside the auto-quote thresholds takes the fast lane. A risk inside appetite but above an authority band goes to the correct senior underwriter. A risk outside appetite goes to decline or referral. Every routing decision carries the inputs and the rationale that produced it.
From that scoring, the layer returns one of three outcomes, each with a written reason. A clean submission that is inside appetite, inside exposure thresholds, and free of fraud flags can receive a quote automatically, which is the fast lane the corretor wants. A case that is clearly outside appetite or hits a hard knockout rule is declined and returned with the reason, so the corretor gets a fast and consistent no instead of silence. A borderline case, one above an authority band, missing data, or carrying a fraud signal, is escalated to a human with all the extracted data and the AI rationale attached.
Escalated cases do not land in an undifferentiated inbox. They enter a prioritized underwriter queue, where priority is computed from factors such as exposure, the win-probability score, expiry urgency, broker value, and the time already elapsed against the SLA. The right underwriter sees the right cases at the top, with context already in front of them. Every submission also carries a visible SLA timer from the moment of intake, so the underwriting team, and where exposed the corretor, can see exactly where a case stands. The net effect is consistency, because the same logic applies to every submission, plus speed for clean business and capacity spent where judgment is actually needed.
How to deploy routing as an external layer
Deploying automatic routing as an external AI layer is a scoped, incremental program, not a core replacement and not a multi-year system overhaul. The contrast is the point. A core replacement is a high-risk, multi-year program that touches the system of record. An external AI layer reuses the existing stack, integrates rather than replaces, and can go live on a defined slice first. With WIR the setup runs 3 to 12 months as a one-time implementation with a fixed price, clear scope, and KPIs agreed before the work starts. Continuous operation follows after go-live, with a billing model adjusted per client.
A pragmatic rollout starts narrow. The first step is scope. Pick one ramo and one channel to begin with, for example patrimonial empresarial through the broker portal, and define the appetite rules, knockout rules, authority bands, and SLA targets that apply. The second step is integration with the core. Connect intake through API, portal, or upload, and connect the write-back to the policy system of record, deciding what is written automatically and what waits for human confirmation. There is no core change. The layer reads and returns.
The third step is the decisive one: calibration to the underwriting manual and risk appetite. The routing and scoring models are tuned to the insurer's own manual, appetite, and historical loss data, never to a generic template, and the thresholds for quote, decline, and escalation are set by the insurer's own risk policy. The fourth step is testing in shadow mode, running the layer in parallel on live submissions and comparing its routing against the underwriters' calls before any automated action goes external. The fifth step is a staged go-live, turning on straight-through processing for the safest band first and widening the auto-quote envelope as audit evidence accumulates. The sixth step is continuous operation, monitoring routing accuracy, false-escalation and false-auto rates, SLA adherence, and model drift, with underwriter overrides feeding back as a training signal. Brazil's structural reason to favor this path is plain. Because 70% of insurers cite IT limitations as the blocker to innovation, the external-layer approach delivers the automation without the core program, and our market-intelligence guide on inteligencia-seguros covers that backdrop in more depth.
Governance, explainability, and LGPD
Every routing and decision the layer makes is explainable and returns a full audit trail, and in the Brazilian frame that is not optional. Each quote, decline, or escalation carries the inputs and the rationale that produced it, so the underwriting team can reconstruct exactly why any case took the path it did. That documented rationale supports both internal governance and the supervision expectations of SUSEP, the supervisor of the private insurance market, around sound and documented underwriting. The audit trail is what separates an auditable decision from a black box.
The data framing follows LGPD, the Lei Geral de Proteção de Dados, Lei 13.709/2018. Insurance submissions carry personal and sometimes sensitive data, so the layer processes it on a lawful basis with data minimization, purpose limitation, and security. Article 20 of the LGPD gives the data subject the right to request review of decisions taken solely on automated processing that affect their interests, which is precisely why an automated routing or underwriting decision needs a documented rationale and a human-review path. The ANPD, the Autoridade Nacional de Proteção de Dados, is the supervisory authority. Data is encrypted at every step, in transit and at rest, across intake, reading, scoring, and write-back.
The model does not impose an external risk view. It encodes the insurer's own appetite and manual, so the insurer remains the risk owner and the decision authority on every referred case. Escalation to a human is a governance feature, not only an operational one, because borderline and high-exposure cases always reach a person. WIR is not an insurer, a broker, or an MGA, and it does not carry risk. It automates the quotation and underwriting journey according to the insurer's own risk-acceptance policy, with decisions that are explainable, auditable, LGPD compliant, and encrypted at every step.
How WIR automates routing and the underwriter queue
WIR is the AI layer for insurance. On top of the systems the insurer already runs, never in their place. It is an external AI layer for insurers and brokers in Brazil that performs the decision-and-prioritization stage described above, calibrated to each insurer's risk appetite and underwriting manual, and it is 100% external with no load on the insurer's IT and no core migration. Two product modules do the work. Underwriter Intelligence automates the quotation journey per the insurer's risk policy, with real-time Machine Learning scoring calibrated to appetite, automatic routing by appetite and exposure, and predictive conversion analysis by product, risk, and broker, so underwriters spend their time on risk judgment and business development rather than triage. Smart Sales is the distribution-side counterpart that maps the portfolio by client and product, scores upsell and next-best-action, and runs multi-channel campaigns with an attribution trail.
In practice, the layer ingests each submission, reads and structures it, scores it against appetite and exposure, and returns a quote, an automatic decline, or an escalation to a human, always with an explanation. It writes the result back to the policy core and returns the audit trail, while a visible SLA and a prioritized underwriter queue keep the team focused on the cases that need judgment. Real-time dashboards, analytics, and reports give a proactive view of in-flight deals and the pipeline. The intelligence is calibrated to the insurer's own underwriting manual, and every automated decision is explainable, auditable, LGPD compliant, and encrypted at every step.
WIR's first public traction is a POC in execution with a global insurer in the Transport line, and the company is conservative about claiming more than that. The reason the external-layer approach resonates is structural. Deloitte puts underwriter time lost to administrative work at 40%, BCG puts insurers blocked from innovation by IT limitations at 70%, Capgemini finds that 60%+ of brokers choose an insurer by response speed, and Gartner estimates that 20-30% of corporate time is lost organizing unstructured data. Automatic routing with a visible SLA is how an insurer answers all four pressures at once. To map how this would prioritize your own underwriter queue, book a conversation with WIR at https://wirinnovation.ai.
Frequently asked questions
How is a submission routed by appetite and exposure?
Each submission is scored on two axes at once: whether the risk sits inside the insurer's defined risk appetite, and where it falls in the exposure and authority band. WIR's Underwriter Intelligence reads and structures the case, scores it with Machine Learning calibrated to the underwriting manual, then sends it down the matching path. A risk inside appetite and inside auto-quote thresholds takes the fast lane, while higher exposure routes to the right senior underwriter, each decision carrying its inputs and rationale.
When does the system quote, decline, or escalate to a human?
The layer returns one of three outcomes, each with a written reason. A clean submission inside appetite, inside exposure thresholds, and free of fraud flags can receive a quote automatically. A case clearly outside appetite or hitting a hard knockout rule is declined and returned with the reason. A borderline case, above an authority band, missing data, or carrying a fraud signal, is escalated to a human with all extracted data and the rationale attached. The insurer remains the risk owner on every referred case.
Does the underwriter see the queue and SLA for each case?
Yes. Escalated cases enter a prioritized underwriter queue rather than an undifferentiated inbox, with priority computed from exposure, win-probability score, expiry urgency, broker value, and time elapsed against the SLA. The right underwriter sees the right cases at the top, context already in front of them. Every submission also carries a visible SLA timer from the moment of intake, so the team, and where exposed the broker, can see exactly where a case stands.
Does automatic routing respect the underwriting manual?
Yes. The routing and scoring models are calibrated to the insurer's own underwriting manual, risk appetite, and historical loss data, never to a generic template. The thresholds for quote, decline, and escalation are set by the insurer's own risk policy, so WIR encodes the insurer's risk view rather than imposing an external one. WIR is not an insurer, broker, or MGA and carries no risk. Every decision is explainable, auditable, LGPD compliant, and encrypted at every step.
How long until routing goes into production?
With WIR, setup runs 3 to 12 months as a one-time implementation with fixed price, clear scope, and KPIs agreed before the work starts. The path is incremental: scope one line and channel, integrate with the existing core through API, portal, or upload, calibrate to the underwriting manual, then test in shadow mode before a staged go-live. There is no core migration. Continuous operation follows after go-live, with a billing model adjusted per client.